I am trying to compare incidence rate ratios (IRR). This is what I have:

Group A (exposed vs. unexposed) IRR and Group B (exposed vs. unexposed) IRR

So how would I correctly test for the difference between the IRR from Group A and the IRR from Group B? The question I would like to answer is whether the incremental differences observed in Group A are similar to those observed in Group B? Please help!

  • 1
    $\begingroup$ Sec. 6.3 of Agresti 2002 (2nd ed.) discusses some options, including the CMH test mentioned by @JamesStanley If you are an R user you may find the R/S companion to this book helpful, specifically p.102 sec E. The R function mantelhaen.test runs the CMH test. $\endgroup$ – conjugateprior Feb 1 '13 at 11:42
  • $\begingroup$ Oops. That was more of an answer than I was intending. Perhaps @James-Stanley wants to incorporate it into his. $\endgroup$ – conjugateprior Feb 1 '13 at 11:44

Generally speaking, you could either apply a Mantel-Haenszel test for whether the impact of the exposure factor differs between groups A and B (basically a stratified chi-squared test), or almost equivalently, model with Poisson regression and include an interaction term between exposure and group status to test whether these differ (EDIT: I mean whether exposure differs by group here too)

Wikipedia is (for once) not very verbose on this subject, but does give a two-liner summary. http://en.wikipedia.org/wiki/Cochran%E2%80%93Mantel%E2%80%93Haenszel_statistics

| cite | improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.